What happened
OpenAI’s Help Center says it will retire Advanced and Standard voice from ChatGPT Voice on December 11, 2026. Users will be notified starting November 10 and can continue using either mode during the transition period. On the retirement date, neither option will remain selectable in Settings > Voice; people who previously selected one will default to Live voice. OpenAI says conversations remain in ChatGPT history and can be continued in text or voice.
The change follows the introduction of GPT-Live in July. OpenAI describes GPT-Live as its latest voice model, combining natural conversation with what it calls frontier intelligence. The company says voice usage has nearly doubled since its introduction and virtually all ChatGPT voice users now use GPT-Live. It says retiring the older options will simplify the experience and let it focus on improving GPT-Live.
OpenAI distinguishes the three voice systems by how they work. Live voice, powered by GPT-Live, is full duplex: it can listen and speak at the same time. It can signal attention with brief phrases, handle quick back-and-forth, or stay quiet while a user thinks. OpenAI also says GPT-Live can delegate complex questions or agentic tasks to its frontier text models, continuing the conversation while background work proceeds.
Advanced voice uses a previous turn-based model that natively processes and generates audio, but waits for the user to stop speaking before replying. Standard voice chains three models: speech-to-text, a language model that produces the response, and text-to-speech. OpenAI says this enabled its original voice experience but led to slower responses and information loss across voice and text modalities.
There are exceptions to the move. Live video and screen sharing will continue to use Advanced voice until GPT-Live supports those capabilities. Voice with custom GPTs will also continue using Advanced voice until its deprecation on December 11, 2026. OpenAI says it is carrying lessons from the older modes into GPT-Live, focusing on response depth and detail, continuity across conversations, and personalization.
What the archive adds
New Runtime’s earlier coverage on GPT-Live voice releases (October 21, 2026) described changes to the user-facing voice layer through model and voice updates. This is a new product change on that known story: OpenAI is now giving two older modes a retirement date and specifying how ChatGPT users will transition.
Trend signal
This supports the direction in Agent-ready software exposes capabilities, but only weakly: GPT-Live’s ability to delegate complex questions and agentic tasks to text models points toward a voice interface that can invoke capabilities beyond conversation. The announcement does not describe the tools, permissions, or observable boundaries involved, so it is not evidence of a mature agent integration surface.
Why now
OpenAI ties the retirement to GPT-Live’s July introduction and says nearly all ChatGPT voice users now use it, while voice usage has nearly doubled. The company presents consolidation as a way to simplify the experience and focus improvement work on GPT-Live.
Why it matters
Teams that depend on ChatGPT voice for customer-facing or internal workflows should treat December 11 as a migration deadline, not assume that a familiar voice option will remain selectable. They should test the replacement in the actual workflow, especially where turn-taking, response style, or continuity matters. The announcement identifies Advanced voice as the temporary path for live video, screen sharing, and custom GPT voice, so workflows using those features do not yet map cleanly to GPT-Live. The broader product signal is that OpenAI is consolidating voice around a full-duplex model that it says can keep a conversation moving while delegating work to text models. That may make voice more capable, but the announcement does not establish that these delegated tasks are available to third-party builders through an API.
Limits and uncertainty
OpenAI does not quantify the reported near-doubling in voice usage, define the period or measurement behind “virtually all,” or provide comparative performance data for GPT-Live versus the retiring modes. It says the model will continue to improve depth, detail, continuity, and personalization, but gives no dates or measurable targets. The announcement also says video and screen sharing will remain on Advanced until GPT-Live has those capabilities, without stating when that transition will happen. These are product claims from OpenAI, not independently verified usage or capability measurements.
Skill to build now
Create a short voice-workflow acceptance test before the retirement date. Run the same realistic prompts through the current selected mode and Live voice, then record whether interruption handling, response timing, continuity between turns, and the expected handoff to text work still meet the workflow’s requirements. Test video, screen sharing, and any custom GPT separately rather than treating a successful basic voice conversation as proof that the entire workflow has migrated. Keep the results with the workflow owner so a product change is visible before it affects users.
Reader takeaway
If a workflow depends on Advanced or Standard voice, test it on Live voice before December 11 and separately check any reliance on video, screen sharing, or custom GPTs. OpenAI has announced a migration, not a like-for-like guarantee: some capabilities remain on Advanced voice during the transition.